Irregularities in field values in MPI=yes case

[palkinev]

Muhamed found a bug, which I think deserves to be mentioned here.

Test case: Tests/Rans/Channel_Re_Tau_590/Rsm No modification is required in the control file. Run ./Process on 4 threads until 600+ timesteps and stop it.

The first result output file produced is chan-ts000600.pvtu (or chan-ts000600.cgns). Use paraview or VisIt post-processing software to visualize the results.

I think each field in this file has an irregularity (a spike in value) near processor buffer zones, but it is most pronounced at "EddyOverMolecularViscosity".